Depth Map Intra Coding Prediction Using Partitioned Sample Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding techniques for depth maps in 3D video coding face challenges in computational complexity and accuracy due to the need for averaging all neighboring samples to determine predicted depth values, especially for larger block sizes, which increases operational complexity and can introduce prediction errors near partition boundaries.

Innovation Solution

The proposed method determines a predicted depth value by analyzing a partitioning pattern to select representative neighboring samples rather than averaging all samples, reducing the number of operations and improving accuracy by avoiding samples near partition boundaries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If all neighboring samples are averaged to determine predicted depth values, then prediction accuracy may be improved, but computational complexity increases significantly

Engineering Contradiction:
Improveprediction accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts only the most relevant neighboring samples (first and last samples of each partition) to form the prediction, discarding the intermediate samples that would otherwise be processed. This selective extraction reduces the number of operations from averaging all samples to averaging just two samples per partition, significantly lowering computational complexity while maintaining prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

2Loss of information

If all neighboring samples are used for prediction, then more information is available, but prediction errors near partition boundaries increase

Engineering Contradiction:
Improveinformation availabilityVSAvoidprediction accuracy
Core Design Contradiction:
Loss of informationVSMeasurement precision

Solution Approach 1:

The patent segments the block of depth values into multiple partitions and applies different prediction strategies to each partition. By using only the first and last samples of each partition for prediction, the method avoids using samples near partition boundaries that would introduce prediction errors, while still maintaining adequate information availability through the segmented approach.

Inventive Principle:
Principle #1Segmentation

3Productivity

If the number of operations is reduced by selecting fewer samples, then computational complexity decreases, but prediction accuracy may deteriorate

Engineering Contradiction:
Improvecoding efficiencyVSAvoidprediction accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by treating different regions (partitions) of the block differently. For each partition, it selectively uses only the boundary samples (first and last) that are most representative of that local region, rather than uniformly processing all samples. This local approach maintains prediction accuracy for each partition while reducing overall computational complexity.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS10375411B2Predictor for depth map intra coding
Publication Date: 2019.08.06 QUALCOMM INC
  • US10375411B2 patent drawing
  • US10375411B2 patent drawing
  • US10375411B2 patent drawing

AI summary

In an example, a process for coding video data includes determining a partitioning pattern for a block of depth values comprising assigning one or more samples of the block to a first partition and assigning one or more other samples of the block to a second partition. The process also includes determining a predicted value for at least one of the first partition and the second partition based on the determined partition pattern. The process also includes coding the at least one of the first partition and the second partition based on the predicted value.